mirror of
https://github.com/alirezarezvani/claude-skills.git
synced 2026-10-10 03:27:56 +00:00
Completes the markdown-html/ domain at 5 skills. The Tier-3 use case from
Shihipar's essay ("Slide Decks"): a markdown deck (slides separated by
--- HR boundaries or # H1 headings, with optional <!-- notes: ... -->
presenter notes blocks) becomes a single-file HTML presentation with
keyboard nav, presenter mode, and print-to-PDF.
Three stdlib tools pipeline together:
1. slide_splitter.py — splits markdown on --- HR or # H1 boundaries
(or --boundary auto: HR wins ≥ 3, else H1 ≥ 5). Extracts the first
heading per slide as the title. Hard rule: refuses 1-slide decks
(exit 5 — it's a poster) and no-boundary input (exit 6 — route to
md-document). Soft-warns slides > 40 source lines (signal-to-noise;
renders anyway).
2. presenter_notes_parser.py — extracts <!-- notes: ... --> blocks
(also speaker-notes: and presenter: aliases) per slide, attaches
as a separate `notes` field, strips from body. Tracks
notes_coverage_pct for the optional --strict-notes gate (refuses
< 50% coverage when presenter mode is essential).
3. deck_html_renderer.py — single-file HTML deck. All slides as
<section class="slide"> elements, one visible at a time (CSS-
controlled). Vanilla JS keyboard handlers: → / Space / PgDn advance;
← / PgUp previous; Home / End first/last; P toggles presenter mode;
Esc exits presenter. URL-hash deep linking (#3 jumps to slide 3,
back/forward walks slides). Progress bar at top (3px); slide counter
bottom-right. Presenter mode = split view: current slide (60% width)
+ panel (40% width with clock + speaker notes + next-slide preview).
@media print { section { display: block; page-break-after: always; } }
→ Cmd+P produces PDF with one slide per page. prefers-reduced-motion
honored throughout. Reuses md-document/scripts/markdown_parser.py
for slide-body content (consistent paragraphs / lists / code / tables
/ callouts). Prism.js is OPT-IN via --syntax (off by default — most
decks don't need it; keeps the file tiny).
Plus 3 references each citing 5-7 sources:
- presentation_ux.md — Atkinson Beyond Bullet Points + Reynolds
Presentation Zen + Tufte Cognitive Style of PowerPoint + NN/g +
Weinschenk + Marp/reveal.js/Big convergence + Tom MacWright
- keyboard_nav_patterns.md — reveal.js/Big/Spectacle keymap + WCAG
2.1.1 + 2.4.3 + MDN KeyboardEvent + NN/g keyboard accessibility
- single_file_deck_conventions.md — Big + Marp + Pandoc + reveal.js
standalone + WCAG 2.3.3 + @media print
1 template asset documenting the canonical single-file deck shape.
/cs:md-slides slash command with 6 pre-flight gates + pipeline +
output digest.
Repo-level updates:
- markdown-html/.claude-plugin/plugin.json: skills array adds
./skills/md-slides; version 2.10.2 → 2.10.3; description marks
domain COMPLETE at 5 skills.
- .claude-plugin/marketplace.json: markdown-html-skills entry version
and description (domain complete); top-level counters 342 → 343
skills, 545 → 548 Python tools, 688 → 691 references, 89 → 90 slash
commands; metadata.version 2.10.2 → 2.10.3.
- Root CLAUDE.md: v2.10.3 release-notes block above v2.10.2.
Validation:
- check_plugin_json.py → OK
- sync-codex-skills.py --dry-run → 1 new symlink, documentation:
5 skills, total 345
- skill_description_validator.py → PASS (all 5 checks: present,
826/1024 chars, third-person, trigger "use after", action verb
"Convert")
- skill_review_checklist_runner.py → 5/6 PASS (under-100-lines warns
at 102; same advisory as md-document SKILL.md)
- All 3 tools pass --help and --sample
- Hard rules verified end-to-end:
no-boundary input → exit 6 with md-document routing hint
1-slide deck → exit 5 with poster recommendation
--strict-notes with < 50% coverage → exit 7
- Full pipeline on 5-slide sample deck (3 with presenter notes)
produces 12.2 KB single-file HTML with all 16 expected components
(slide-1 + slide-5 anchors, notes attribute populated, P-key
handler, arrow nav, @media print, page-break-after, presenter
panel + clock + next-preview, palette tokens, progress bar, title
in header, "1 / 5" counter, history.replaceState URL hash sync,
prefers-reduced-motion).
Domain status: COMPLETE. All 5 planned skills shipped across 4 PRs
(#780 foundation, #793 md-document, #795 md-review, this PR md-slides).
The markdown-html/ domain operationalizes Shihipar's central claim —
markdown collapses past 100 lines; HTML restores density, clarity,
shareability, and lightweight interaction — across all three layout
families (long-form documents, code reviews, slide decks).
https://claude.ai/code/session_01BK2KoQot1U7J5oSosrCQdc
4.1 KiB
4.1 KiB
| description | argument-hint |
|---|---|
| Convert a markdown deck (slides separated by --- HR boundaries or by | <path to markdown deck> [--boundary auto|hr|h1] [--title "My Talk"] [--syntax] [--strict-notes] |
/cs:md-slides — Markdown deck → single-file HTML presentation
Convert the markdown deck at $ARGUMENTS into a single-file interactive HTML presentation.
Pre-flight gates (refuse, never override)
- Input < 100 lines → refuse (markdown wins below Shihipar's threshold).
- Design-system not onboarded → refuse, surface
/cs:design-system. - No clear slide boundaries (auto mode: need ≥ 3 HR or ≥ 5 H1) → refuse, route to md-document.
- 1-slide deck → refuse (it's a poster, not a deck).
--strict-noteswith < 50% notes coverage → refuse.- Output directory unwritable → refuse, ask user for
--out.
Pipeline
# 1. Resolve output path (doctype=slides → deck- prefix)
python3 markdown-html/skills/markdown-html-orchestrator/scripts/output_path_resolver.py \
--input "<path>.md" --doctype slides
# 2. Split slides on --- HR or H1 (auto-detect)
python3 markdown-html/skills/md-slides/scripts/slide_splitter.py \
--input "<path>.md" --boundary auto --output /tmp/slides.json
# 3. Extract <!-- notes: ... --> blocks per slide
python3 markdown-html/skills/md-slides/scripts/presenter_notes_parser.py \
--slides /tmp/slides.json --output /tmp/deck.json
# 4. Render single-file HTML deck
python3 markdown-html/skills/md-slides/scripts/deck_html_renderer.py \
--slides /tmp/deck.json --title "<deck title>" \
--output <resolved-out>.html
What ships in the HTML
- All slides as
<section class="slide">with one visible at a time (CSS-controlled, no JS-required content) - Keyboard navigation:
→/Space/PgDn→ next slide←/PgUp→ previous slideHome/End→ first / last slideP→ toggle presenter modeEsc→ exit presenter mode
- Presenter mode — split view: current slide (60% width) + panel (40% width with clock + speaker notes + next-slide preview)
- URL-hash deep linking —
#3jumps to slide 3; back/forward walks slides; sharedeck.html#5to land on slide 5 - Progress bar at top (3px); slide counter in bottom-right
- Print-to-PDF via browser's native print dialog:
@media printmakes each slide one page (Cmd+P/Ctrl+P) prefers-reduced-motionhonored- 12 brand CSS tokens from design-system; design_style affects layout density
Hard rules
- Output is one
.htmlfile. No multi-file output. - External CDN:
fonts.googleapis.comalways;cdn.jsdelivr.net(Prism) only when--syntaxis passed. - No JS framework runtime. Vanilla JS + keyboard event handlers.
- Re-running on the same input writes
deck-{slug}-2.htmletc.
Useful flags
--boundary {auto,hr,h1}— slide boundary mode (default: auto)--title "My Talk"— sets the<title>and tab name--syntax— enable Prism.js CDN for code blocks (off by default; decks rarely need it)--strict-notes— refuse if < 50% of slides have presenter notes (use when presenter mode is essential)
Output
Returns: slide count, notes coverage %, output path, design style applied, top features used, one forcing question.
References
See markdown-html/skills/md-slides/references/:
presentation_ux.md— Atkinson + Reynolds + Tufte + NN/g + Weinschenk + Marp/reveal.js/Big convergencekeyboard_nav_patterns.md— reveal.js / Big / Spectacle keymap + WCAG 2.1.1 + 2.4.3 + MDN KeyboardEventsingle_file_deck_conventions.md— Big + Marp + Pandoc + WCAG 2.3.3 + @media print